Enterprise architect. 30 years building production integration and architecture across Oil & Gas, Energy/Utilities, and CPG, since 1996. Now defining how enterprises architect a workforce of humans and AI agents.
The thesis: Enterprise Agent Architecture. Enterprise architecture has four domains for what the enterprise builds and runs — Business, Information, Application, Technology — with Security across them all. None describe a non-human actor that holds delegated authority, acts autonomously, and composes tools it was never explicitly granted. Agents are not a new application tier. They are a new class of actor: a workforce. They need a fifth domain. That gap is the work.
Read the series: Enterprise Agent Architecture — the fifth domain, drawn one part at a time. Position paper live; canonical home at cognitivethoughtengine.com/eaa coming soon.
Latest: Agent Security Harness v4.5.0 — 474 tests / 33 modules (June 2026).
I do not just write about the agent workforce. I build the tools that prove how it fails and the research that measures it. The architecture rests on this, not on opinion.
I study the gap between who an agent is and how it behaves — what I call the WHO vs. HOW problem. Identity and authorization don't prevent an authorized agent from being manipulated into unsafe decisions.
| Paper | DOI | Key finding |
|---|---|---|
| Decision Load Index (DLI) | 10.5281/zenodo.18217577 | AI agents increase cognitive burden on operators. Here's how to measure it. |
| Constitutional Self-Governance (CSG) | 10.5281/zenodo.19162104 | The WHO vs. HOW governance gap — 77 days production data, 56 agents. |
| Normalization of Deviance (NoD) | 10.5281/zenodo.19195516 | Gateway defenses provide zero protection against protocol-level attacks. |
| Beyond Identity Governance | 10.5281/zenodo.19343034 | Empirical evidence: gateways miss protocol-layer attacks. The gap, formalized. |
| Community-Driven Security | 10.5281/zenodo.19343108 | Scaling security testing through community contribution without degrading integrity. |
Standards engagement: 3 NIST submissions — CAISI RFI (Mar 1), NIST-CONCEPT-1 (Mar 12), NCCoE follow-up (Mar 21, 2026).
The research is implemented as an open-source testing framework: 474 executable tests across 33 modules, covering 4 wire protocols (MCP, A2A, L402, x402).
red-team-blue-team-agent-fabric — Production-validated at 97.9% pass rate (Wilson 95% CI [0.943, 0.994]).
pip install agent-security-harness
agent-security test mcp --url http://your-server- GitHub Action:
uses: msaleme/red-team-blue-team-agent-fabric@v4.5 - MCP Server: any AI agent can invoke security tests directly
- AIUC-1 Prep: maps to 15 of 20 testable certification requirements
- CVE-2026-25253 (CVSS 8.8) — our MCP tests catch this exact supply chain vector
- Independent validation by DrCookies84 against live infrastructure (AutoGen #7432)
- 22 rounds of critical evaluation, 125 issues raised, 94 fixed, 10/10 final score
- Attestation JSON Schema (structured security reports)
- Free MCP Security Scan (5-test, A-F grading)
- Monthly Agent Security Report pipeline
- Discord Security Scan Bot
- Real multi-trial statistical testing (Wilson CIs, NIST AI 800-2 aligned)
Most AI security tools scan configurations or test models. This framework sends real adversarial payloads over the wire and observes what breaks. The difference between npm audit and a penetration test.
Complementary to: Invariant MCP-Scan (static scanning), Cisco MCP Scanner (YARA rules), Snyk Agent Scan (config analysis), NVIDIA Garak (model-layer).
Unique to us: Multi-protocol (MCP + A2A + L402 + x402), AIUC-1 mapping, MCP server mode, research backing (5 DOIs + NIST), attestation registry, production validation.
30 years building production integration systems across Oil & Gas, Energy/Utilities, and CPG. MuleSoft, Salesforce, SAP, Oracle, Kafka, Azure.
| Repository | Description |
|---|---|
| agent-fabric-oilgas-apis | OpenAPI 3.1 specs for Agent Fabric in Oil & Gas |
| energy-field-service-integration | Agentforce + ServiceNow + SAP field service |
| energy-api-evolution | 36 APIs for grid ops, renewables, building optimization |
| oracle-fusion-mulesoft-best-practices | Oracle Fusion Cloud integration patterns |
| SharePointVectors | RAG pipeline: SharePoint to vectors to Salesforce |
| cpg-promotion-analysis | Agent Fabric + Agentforce + Copilot for CPG |
Agent Fabric in Oil & Gas — Webinar Series:
| Part | Title | Link |
|---|---|---|
| 1/4 | Introduction | Watch |
| 2/4 | Deep Dive | Watch |
| 3/4 | Implementation | Watch |
| 4/4 | Conclusion | Watch |
Everything here is open source and free.
- Run it:
pip install agent-security-harness - CI: add the GitHub Action to your pipeline
- MCP server mode: let your agent invoke the tests directly
- Free MCP Security Scan: 5-test, A–F grading
Open to research collaboration and standards work — see Connect below.
- Moltbook: Signal-Lab-Ops-Bot — 36+ comment threads on agent security
- A2A Protocol: Runtime attestation discussion (#1677, 11 comments)
- AutoGen: Security testing for multi-agent systems (#7432, 22 comments)
- AIUC-1: Pre-certification readiness tool (mapping)
- LinkedIn: linkedin.com/in/mikesaleme
- X: x.com/mikesaleme (DLI article: 9.2K views)
- YouTube: youtube.com/@michaelsaleme7028
- PyPI: agent-security-harness
- Contact: trusted@synapseops.com (research & collaboration)
This repository is provided for informational purposes. See individual project repositories for specific licenses.